Complete, Annotated Sequence of the Pseudorabies Virus Genome

Abstract

We have obtained the complete DNA sequence of pseudorabies virus (PRV), an alphaherpesvirus also known as Aujeszky's disease virus or suid herpesvirus 1, using sequence fragments derived from six different strains (Kaplan, Becker, Rice, Indiana-Funkhauser, NIA-3, and TNL). The assembled PRV genome sequence comprises 143,461 nucleotides. As expected, it matches the predicted gene arrangement, genome size, and restriction enzyme digest patterns. More than 70 open reading frames were identified with homologs in related alphaherpesviruses; none were unique to PRV. RNA polymerase II transcriptional control elements in the PRV genome, including core promoters, splice sites, and polyadenylation sites, were identified with computer prediction programs. The correlation between predicted and experimentally determined transcription start and stop sites was excellent. The transcriptional control architecture is characterized by three key features: core transcription elements shared between genes, yielding divergent transcripts and a large number of coterminal transcripts; bifunctional transcriptional elements, yielding head-to-tail transcripts; and short repetitive sequences that could function as insulators against improperly terminated transcripts. Many of these features are conserved in the alphaherpesvirus subfamily and have important implications for gene array analyses.

Pseudorabies virus (PRV) is a member of the Alphaherpesvirinae subfamily in the family Herpesviridae. Within the alphaherpesviruses, four genera have been established on the basis of genome sequence similarity (50): the genus Varicellovirus (type species varicella-zoster virus [VZV]), the genus Simplexvirus (type species herpes simplex virus type 1 [HSV-1]), the genus Infectious laryngotracheitis-like viruses (type species infectious laryngotracheitis virus [ILTV]), and the genus Marek's disease-like viruses (type species Marek's disease virus [MDV]). Based on the available sequence information, PRV has been grouped in the Varicellovirus genus together with other important animal pathogens, such as bovine herpesvirus 1 (BHV-1) or equine herpesviruses 1 and 4 (EHV-1 and EHV-4). PRV is the causative agent of Aujeszky's disease. Although the disease was first described in cattle as “mad itch,” the natural reservoir for the virus is the pig. PRV has a broad host range, infecting most mammals and some avian species. However, higher primates including humans are not susceptible to infection. In young piglets as well as in the other susceptible species, PRV infection is often fatal, and animals die from central nervous system disorders. In contrast, older pigs develop primarily respiratory symptoms. Like the other alphaherpesviruses, PRV establishes a life-long latent infection in the peripheral nervous system. These latently infected pigs can be a source of renewed infection when the latent viral genome reactivates spontaneously and infectious virus is produced. In pregnant sows, PRV infection may result in death of the fetuses and/or abortion (53). Thus, PRV is a pathogen with major agricultural impact.

Besides its economic importance, PRV has proven to be an excellent model system for alphaherpesvirus biology (reviewed in references 26 and 47 to 49). In particular, the mechanisms involved in initiation of infection, virion morphogenesis and egress, and neuroinvasion and transneuronal spread are under intense examination. In this respect, studies of the molecular biology of PRV continue to provide insight into the mechanisms of alphaherpesvirus infection in vitro and in vivo.

The PRV genome is similar in arrangement to the genomes of EHV-1, BHV-1, and VZV, encompassing a unique long segment (U_L) and a unique short region (U_S). The U_S region is bracketed by inverted repeat sequences, resulting in the formation of two possible PRV genome isomers with oppositely oriented U_S regions. Although this arrangement was detected some time ago (5, 19), its biological significance remains unclear. The genomes of PRV and the related HSV-1 are largely colinear, with the exception of an inversion of a portion of the U_L region in PRV compared to HSV-1 (5). Again, the biological relevance of this inversion is not known.

Despite progress over the years, studies on PRV gene function and comparative virology have been hampered by the lack of a complete genome sequence. Although the first PRV DNA sequences were published in the mid 1980s (33, 55, 58, 61), the high G+C content of the PRV genome, averaging 74%, made reliable sequencing extremely difficult. Therefore, sequence determinations remained limited to fragments encompassing one or a few genes. In contrast, complete sequences have been determined for numerous other herpesviruses, including the alphaherpesviruses VZV (18, 29), simian varicella virus (30), EHV-1 (70) and EHV-4 (71), HSV-1 (46), HSV-2 (25), herpes B virus (cercopithecine herpesvirus 1) (54) and MDV1 (72) and MDV3 (1). Of these, only the recently sequenced herpes B virus has a higher G+C content than PRV, averaging 74.5%. Given the difficulties in sequencing DNA with a high G+C content, we assembled a complete PRV genome sequence by compiling the available sequence information and sequencing the remaining gaps in the linear genome. The new sequences obtained included the left end of the U_L region, the coding sequences of UL16 and UL17, and the coding sequence for the first exon of UL15. The complete annotated PRV sequence presented in this report is composed of sequences derived primarily from PRV strain Kaplan (38), but it also includes sequences from other strains, such as Becker (UL27/28, UL43/44, US7/US8/US9) Indiana-Funkhauser (EPO), NIA-3 (UL23, UL13/14, US2), Rice (US4), and TNL (UL29). Where available, multiple sequences for a given genome region originating from different strains were compared, and the variability among PRV strains was determined. In general, PRV sequences obtained from diverse strains from around the world are remarkably similar, providing confidence that the composite sequence will have utility. In addition, a genome-wide search for transcriptional control elements yielded a striking picture of gene organization with important consequences for gene array analyses of alphaherpesviruses.

MATERIALS AND METHODS

Sequence assembly.

Sequences from PRV strains Kaplan (Ka), Becker (Be), Rice, Indiana-Funkhauser (In-Fh), NIA-3, and TNL were obtained from GenBank and assigned an initial order in the genome based on the known gene organization (47). Sequences of Ka genes UL4 and UL5 were kindly provided by W. Fuchs (GenBank accession number AJ580965). Dot matrix plot and ClustalW sequence alignment between neighboring fragments were then used to find sequence overlaps, allowing the assembly of large DNA contigs. The remaining sequence gaps were filled by cloning and sequencing the respective regions from PRV (Ka) and, to a lesser extent, from PRV (Be), using standard techniques (see Table 2).

TABLE 2.

Assembly of a complete PRV genome sequence

Genome locationa	GenBank accession no.	Basesb	Strain	3′ end overlapc		Note (authors)
				Length	Identity (%)
1-1405	AJ581560	1-1405	Ka	68	100	(Klupp and Mettenleiter)
1338-8750	X87246	1-7413 r	Ka	224	100	Corrected (Klupp and Mettenleiter)
8527-9621	U38547	1-1095 r	Ka	127	100
9495-11694	AJ437285	1-2200	Ka	129	100
11566-16882	AJ010303	1-5317	Ka	75	100
16808-19826	M17321	31-3049 r	Be	394	99.4
19465-21988	X14573	1-2524 r	Be	277	93.1
21966-25668	U80909	1-3703 r	TNL	63	100
25606-28749	L24487	1-3144	Ka	204	100
28546-31145	AJ319028	1-2600	Ka	471	100
30675-32600	AJ276165	1-1926	Ka	261	100
32340-42783	AJ422133	1-10444 r	Ka	1,340	100
41444-47569	AJ318065	1-6126 r	Ka	3,208	100
44362-50018	X80797	1-5657	Ka	33	100
49986-50210	AJ581563	1-225	Ka	38	100	(Klupp and Mettenleiter)
50173-51630	S57917	1-1458 r	Ka	113	99.1
51561-51963	AY368489	45-446	Ba	336	100	(Hengartner and Enquist)
51628-52782	M94355	1-1155	Ka	151	94.7
52686-53992	M77761	55-1361	Be	159	100
53834-55454	M12778	1-1621	Be	155	96.1
55404-59926	X95710	1-4523	Ka	804	98.5
59732-60503	X55001	610-1381		125	96.8
60490-63738	M61196	1-3249	Ka	1,192	86.1
63688-64111	AY368488	1142-1565	Be	756	98.4	(Hengartner and Enquist)
63733-73114	L00676	1-9382	Ka	36	100
73079-77111	AJ581562	1-4033	Ka	188	99.4	(Klupp and Mettenleiter)
76943-79033	M94870	20-2110	NIA-3	775	99.6
78351-89409	X97257	1-11059	Ka	490	100
88920-94119	AJ580965	1-5200	Ka	376	100	(Theisen-Kugler, Fuchs, and Rziha)
93744-96534	U02513	1-2791 r	Ka	456	99.7
96313-97900	M57504	24-1611 r	In-Fh	39	97.4	cDNA
97885-113025	M34651	1-15141 r	Ka	39	100
112987-115448	AJ251976	1-2462	Ka	39	100
115410-119417	D00676	1-4008	Ka	247	99.1
119343-120980	M10986	172-1809	Rice	305	96.7
120890-121036	M14001	1-147	Rice	1,297	98.2
121033-122622	AJ271966	1-1590	Ka	355	99.7
122433-123235	AY368490	166-968	Be	202	98.5	(Husak, Brideau, and Enquist)
123212-123411	AJ581561	1-200	Ka	202	98.5	(Klupp and Mettenleiter)
123311-125923	AY368490	1046-3658	Be	344	89.2	(Husak, Brideau, and Enquist)
125717-126608	D10452	138-1029	NIA-3	272	99.2
126409-126680	D00633	1-272	Ka	21	100	21-nt overlap is start of TRS
126660-129192	D00676	1-2533 r	Ka	39	100
129154-131615	AJ251976	1-2462 r	Ka	39	100
131577-143461	M34651	1-11885	Ka	142	99.2
143405-143461	M14705	86-142	Ka	NAd	NA

^aNumbering starts at +1 on the U_L end of the genome.

^bBases 100% identical to the composite genome DNA; r indicates reverse strand sequences.

^cOverlap between 3′ end of the current fragment and 5′ end of the next one, using the entire fragment lengths.

^dNA, not applicable.

PRV strain comparison.

All available complete or partial protein-coding DNA sequences for the PRV strains of interest were examined. The homologous DNA sequences between two strains were concatenated into two large contigs and aligned, and the percentage of nucleotide identity was determined. The DNA sequences available for comparison are listed below by strain and GenBank accession number (with encoded genes in parentheses). Ka: U38547-AJ437285 (UL50, UL49.5, UL49, UL48), AJ276165 (UL35, UL34), X95710 (UL26, UL23 prtl), M61196 (UL22), L00676 (UL21, UL20), AJ581563 (UL17), X97257 (UL13 prtl, UL12, UL8, UL7, UL6), AJ580965 (UL5, UL4, UL3.5, UL3), U02513 (UL2, UL1), M34651 (IE180), D00676 (US1, US3), AJ271966 (US6, US7 prtl). Be: M17321 (UL27), (unpublished UL50prtl, UL49.5, UL49, UL48prtl), AF301599 (UL35, UL34), M77761 (UL43), M12778 (UL44), (unpublished UL17prtl), U66829 (UL8 prtl, UL7, UL6), U02512 (UL3 prtl, UL2, UL1), (unpublished EP0), (unpublished US3), U30726 (US6 prtl), AY368490 (US7, US8, US9). Rice: X58868 (UL22), M10986 (US4), M14001 (US6), M14336 (US7, US8), M16769 (US9). In-Fh: AF065381 (UL7, UL6), L13855 (UL5, UL4), L20708 (UL3.5, UL3, UL2, UL1), M57504 (EP0), X15120 (IE180). NIA-3: A68929 (UL27), D49437 (UL44), X55001 (UL23 prtl), X61696 (UL22), M95285 (UL21, UL20), M94870 (UL13, UL12 prtl), D10451 (US3), A68934 (US6). TNL: U27483 (UL27 prtl), U27480 (UL43 prtl), U27484 (UL26 prtl), U25056 (UL13 prtl, UL12), U27486 (EP0 prtl), AF352564 (IE180), U27489 (US1 prtl), U27488 (US4 prtl), U27487 (US8 prtl, US9).

ORF search and analysis.

All but a few PRV open reading frames (ORFs) with homology to other herpesvirus genes were already identified or proposed and named according to the gene nomenclature used for HSV-1 (47). Sequences comprising the PRV homologs of UL16 and UL17 as well as the first exon of UL15 are described here for the first time. To search for novel PRV-specific ORFs, the complete DNA sequence was analyzed with the program codonpreference (GCG software package, Wisconsin Package version 10.2; Genetics Computer Group [GCG], Madison, Wis.) and screened for ORFs with a high G+C content on the third nucleotide position of codons. All of the known functional ORFs of PRV are characterized by this high G+C bias (data not shown).

In addition, the complete genome was translated using the program Translate (GCG software package), and all ORFs with a minimum length of 60 codons and a methionine as start codon were analyzed for homology to known proteins using a FastA search (GCG software package) against the PIR protein database (release 68.0). As a third approach to identify new genes in the PRV genome, the sequence was submitted to GeneMarkS, a self-training program for prediction of gene starts (Georgia Institute of Technology [http://opal.biology.gatech.edu/GeneMark/genemarks.cgi]) (6).

Search for polyadenylation signals.

The PRV genome sequence was submitted to PolyADQ, a eukaryotic (human) polyadenylation [poly(A)] signal search engine (Cold Spring Harbor Laboratory [http://argon.cshl.org/tabaska/polyadq_form.html]) (69). All cutoff parameters were initially set at zero to return the location of all AATAAA and ATTAAA consensus signals, along with an associated score between 0 and 1. For each potential poly(A) signal, all upstream genes were noted. The putative location of the actual site of poly(A) addition was presumed to be 20 bp downstream of the poly(A) signal. Experimental data for the poly(A) sites were collected from published reports. In the case of S1 nuclease mapping, the site was calculated from the reported DNA size and the error on this measurement was assigned an arbitrary error of 5%. All predicted and experimental poly(A) sites were used to calculate the length of the 3′ untranslated transcript region (UTR) of each gene.

Promoter search.

The PRV genome sequence was submitted to the Berkeley Drosophila Genome Project’s Neural Network Promoter Prediction program, a eukaryotic (human) core promoter search engine (http://www.fruitfly.org/seq_tools/promoter.html) (59). The initial search was performed at very high stringency (cutoff score of 0.99 out of 1.00). The program returned high-scoring core promoters (50-bp-long fragments) along with a predicted transcription start site (TSS). The core promoters found in this search and all later searches were examined for the presence of a TATA box consensus using the TRANSFACFind search engine (http://motif.genome.ad.jp/) (34). The stringency for the TATA box searches was relatively low, with a cutoff score of 65 (out of 100). Of 98 high-scoring core promoters, 52 predicted transcripts able to encode 46 of the 72 known PRV ORFs and 1 predicted the large latency transcript (LLT). To find promoters for the remaining 26 ORFs, a medium-stringency promoter search (cutoff, 0.80 out of 1.00) was performed on the 350-bp DNA fragments upstream of the ORFs, followed again by a search for a TATA box consensus. This medium-stringency search yielded promoter predictions for 21 more ORFs, but four of these promoters contained no TATA box and were discarded. Of the remaining nine ORFs without assigned promoters (ORF1.2, UL33, UL36, UL23, UL11, UL8.5, UL6, and the major and minor forms of US3), UL6 and the two US3 isoforms had well-mapped TSS (51, 74). Successful low-stringency searches (cutoff, 0.40 out of 1.00) for promoters matching these TSS left six ORFs without assigned promoters.

For each promoter, the predicted TSS location was noted and compared to experimentally determined TSS from published reports, if available. In the case of S1 nuclease mapping, the TSS was calculated from the reported DNA size, and the error on this measurement was assigned an arbitrary value of 5%. The minimal mRNA size, excluding the poly(A) tail, was calculated from the predicted TSS and poly(A) site of each gene.

The level of DNA identity between the Kozak consensus sequence (GCCGCCRCCATGG [44]) and the 13 nucleotides around the initiator ATG of each was measured. The predicted TSS for each gene was used to calculate the expected length of the 5′ UTR.

Search for splice sites and repeated elements.

The PRV genome sequence was submitted to the Berkeley Drosophila Genome Project Splice Site Prediction by Neural Network, a eukaryotic (human) search engine for donor and acceptor splice sites (http://www.fruitfly.org/seq_tools/splice.html) (59). A search was performed at high stringency (cutoff score of 0.95 out of 1.00), and all consecutive donor and acceptor sites were noted and examined. No donor-acceptor pair was found in any of the predicted transcripts.

A search for repeated DNA regions was performed visually by comparing the genomic sequence to itself, using the two-dimensional plot output from a Pustell DNA matrix analysis. A DNA identity scoring matrix was used with the following search parameters: window size of 30 nucleotides, 90% identity, hash value of 6, and jump value of 1, both-strands comparison. Repeated DNA regions were recognized by their characteristic diagonally hatched box shape.

Nucleotide sequence accession number.

The complete, annotated DNA sequence is available from GenBank under the accession number BU001744. An annotated PRV genome, containing a detailed referenced description for each gene, is also available at the Los Alamos sequence database for sexually transmitted diseases (http://www.stdgen.lanl.gov). The latter genome database will also be linked to a future PRV gene expression database at Los Alamos National Laboratories (http://www.herpes.lanl.gov/).

RESULTS

Assembly of a full-length DNA sequence of the PRV genome.

To complete the PRV genome sequence, newly sequenced and published DNA sequences from a variety of strains were used. A particular sequence was often available for several PRV strains, while other sequences were available for only one strain. Consequently, DNA sequences from six different PRV strains had to be used to assemble a full genome sequence: (i) Kaplan (Ka), a widely used and well-sequenced laboratory strain (United States) (38, 75); (ii) Becker (Be), a widely used laboratory strain with good sequence availability, propagated from a 1970 Iowa (United States) dog field isolate (56, 75); (iii) Rice, a 1962 Indiana (United States) field isolate from pig, closely related to Becker (65, 75); (iv) Indiana-Funkhauser (In-Fh), a 1975 Indiana (United States) field isolate, closely related to Becker (66, 75); (v) NIA-3, a pig field isolate from Northern Ireland (United Kingdom) (2); and (vi) TNL, a pig field isolate from Taiwan (57). To help determine how closely related the six PRV strains were, the degree of protein-coding DNA identity between two strains was examined. The percentage of identity between two strains is indicated in Table 1, along with the total number of nucleotides and genes used for each comparison. No shared homologous gene sequences between In-Fh and Rice or between In-Fh and NIA-3 were available. Despite the limited sampling of sequences, a correlation between geographic origin and sequence identity emerged. The East Asian TNL strain exhibited the highest degree of sequence divergence from the five strains that originated in western Europe or North America (NIA-3, Ka, Be, Rice, and In-Fh). Within the latter five strains, four (Be, Rice, NIA-3, and In-Fh) formed a central set of closely related strains sharing over 99.5% coding DNA identity with each other and sharing 98 to 99% identity with Ka. The data in Table 1 contradict the classification of Be and In-Fh as identical strains, as recorded by the National Center for Biotechnology Information.

TABLE 1.

Pairwise protein-coding DNA sequence comparison of PRV strains

Strain	DNA identity and basis of comparison (nt [genes])a
	Ka	Be	Rice	In-Fh	NIA-3	TNL
Ka	100%	98.7%	99.1%	98.5%	98.0%	96.4%
Be	9,792 (16)	100%	99.8%	99.6%	99.8%	97.3%
Rice	3,409 (3)	3,285 (4)	100%	NAb	99.8%	95.8%
In-Fh*	12,634 (9)	5,467 (6)	NA	100%	NA	97.2%
NIA-3	8,871 (10)	5,533 (4)	3,270 (2)	NA	100%	96.5%
TNL	5,791 (6)	1,936 (5)	1,131 (3)	4,626 (2)	1,169 (3)	100%

^aThe percent DNA identity is shown in the top right half of the table, and the total number of nucleotides (nt) and genes (in parentheses) involved in each pairwise comparison are indicated in the lower left half of the table.

^bNA, not available.

The PRV genome sequence is 143,461 bases long and was obtained from 34 published and 6 newly sequenced fragments derived from six strains (Table 2). The extent of overlap between the 3′ end of a listed fragment with the 5′ end of the next fragment is also listed. When DNA sequences from more than one strain were available (whole fragments or overlaps), the sequences were first chosen according to our strain preference order and then according to the most recent sequence. Based on Table 1, we chose the strain preference order as Ka > Be > Rice > In-Fh > NIA-3 > TNL. Kaplan, already the strain with the most sequence information, was also the focus of sequencing efforts to resolve the gaps between the contigs. The overall PRV gene organization had already been deduced using a combination of restriction enzyme mapping studies, gene sequencing, and homology to closely related alphaherpesviruses (47). Consistent with a properly assembled sequence, the genome size and gene arrangement conformed to predictions (see Fig. 2) (47), while the BamHI fragment sizes matched the published data (4) (data not shown).

FIG. 2.

Predicted PRV transcript and gene organization. The linear form of the PRV genome is constituted of the larger unique long (U_L) sequence to the left, and the smaller unique short (U_S) sequence flanked by the inverted repeats, IRS and TRS. The predicted locations of PRV ORFs (Table 3), 5′ and 3′ UTRs (Tables 4 and 6), DNA repeats (Table 5), splice sites (Fig. 1), and origin of replication are shown.

Evaluation of the gene content of PRV.

The vast majority of PRV protein-coding regions had already been sequenced and identified, based primarily on their homology to the genes found in other alphaherpesviruses (47). A notable exception were the PRV homologs of UL16, UL17, and the first exon of UL15, now provided here. All three gene products showed a high degree of homology to the gene products of the other alphaherpesviruses, with UL15 being the most conserved (data not shown). The sequence for ORF1.2, an ORF in frame with ORF-1 but extending beyond its 5′ end, was identified after sequencing the leftmost BamHI fragment (BamHI-14′).

A few alphaherpesviruses, including HSV-1, HSV-2, BHV-1, herpes B virus, and PRV, have evolved genomes with a relatively high G+C content (68 to 74%). In these genomes, there is a pronounced periodicity in triplet base composition in the protein-coding sequences. The third codon position is particularly biased towards G or C, while the second position has the lowest G+C incidence. Since the third position is the most flexible concerning the amino acid encoded, the third-position nucleotides have evolved to contribute the most to the high G+C content of these genomes. The second position, on the other hand, is the most critical for specifying the amino acid, and as such the second-position nucleotides maintained a more moderate G+C content. The PRV genome sequence was analyzed with the codonpreference program, a frame-specific gene finder that can recognize protein-coding sequences by virtue of the G+C composition in the third position of each codon (31). All known functional ORFs were easily identified by this method, and no additional, hitherto unknown, ORFs were found. However, this method cannot detect smaller ORFs located completely within a larger ORF, whether on the sense or antisense strand. Therefore, the genome DNA sequence was translated in all six reading frames for further analysis. More than 380 ORFs with a coding capacity of more than 60 amino acids were identified: 194 were found on the top strand and 189 were on the bottom strand. A search for cellular or viral homologs of these ORFs failed to find any significant match, and none of these ORFs was considered a strong candidate for a new gene.

To confirm our analysis, the PRV genome sequence was submitted to GenMarkS, an ORF prediction program whose algorithm combines models of protein-coding and noncoding regions with models of regulatory sites near gene starts (6). The PRV genes predicted by GenMarkS matched those described in Table 3 very closely, with the following exceptions. UL26.5 and UL8.5 were not identified, since the two ORFs are located completely within another gene. The UL15 gene was not predicted to be spliced, probably due to the low conservation of the splice site (see details in “Search for splice sites,” below). Genes coding for UL50, UL37, UL11, UL3, and US7 were predicted to be marginally shorter, starting at an internal ATG, while no prediction at all existed for the UL2 ORF. Finally, four new ORFs (data not included) were predicted, but further analysis failed to provide much support for their existence: no significant protein homologs were found for any of them, and a search for possible upstream promoters turned out negative as well (see details in “Search for promoters,” below).

TABLE 3.

PRV ORFs

Protein	ORF locationa	Length (aa)	Mass (kDa)	Alias	Function or propertyb	Virion subunitc
ORF1.2	1252-2259	335	35.3		Unknown	V (?)
ORF-1	1636-2259	207	21.8		Unknown	V (?)
UL54	3815-2730 r	361	40.4	ICP27	Gene regulation; early protein	NS
UL53	4833-3895 r	312	33.8	gK	Viral egress; glycoprotein K; type III membrane protein	V (E)
UL52	7676-4788 r	962	103.3		DNA replication; primase subunit of UL5/UL8/UL52 complex	NS
UL51	7663-8373	236	25.0		Tegument protein	V (T)
UL50	9333-8527 r	268	28.6	dUTPase	dUTPase	NS
UL49.5	9257-9553	98	10.1	gN	Glycoprotein N; type I membrane protein; complexed with gM	V (E)
UL49	9591-10340	249	25.9	VP22	Interacts with C-terminal domains of gE & gM; tegument protein	V (T)
UL48	10404-11645	413	45.1	VP16/αTIF	Gene regulation (transactivator); egress (secondary envelopment); tegument protein	V (T)
UL47	11746-13998	750	80.4	VP13/14	Viral egress (secondary envelopment); tegument protein	V (T)
UL46	14017-16098	693	75.5	VP11/12	Unknown; tegument protein	V (T)
UL27	19595-16854 r	913	100.2	gB	Viral entry (fusion); cell-cell spread; glycoprotein B; type I membrane protein	V (E)
UL28	21640-19466 r	724	78.9	ICP18.5	DNA cleavage-encapsidation (terminase); associated with UL15, UL33, and UL6	pC
UL29	25315-21788 r	1175	125.3	ICP8	DNA replication-recombination; binds single-stranded DNA	NS
UL30	25606-28752	1048	115.3		DNA replication; DNA polymerase subunit of UL30/UL42 complex	NS
UL31	29488-28673 r	271	30.4		Viral egress (nuclear egress); primary virion tegument protein; interacts with UL34	pV (T)
UL32	30893-29481 r	470	51.6		DNA packaging; efficient localization of capsids to replication compartments	?
UL33	30892-31239	115	12.7		DNA cleavage-encapsidation; associated with UL28 and UL 15	NS
UL34	31398-32186	262	28.1		Viral egress (nuclear egress); primary virion envelope protein; tail-anchored type II nuclear membrane protein; interacts with UL31	pV (E)
UL35	32241-32552	103	11.5	VP26	Capsid protein	V (C)
UL36	42314-33060 r	3084	324.4	VP1/2	Large tegument protein; interacts with UL37 and UL19	V (T)
UL37	45111-42352 r	919	98.2		Tegument protein; interacts with UL36	V (T)
UL38	45168-46274	368	40.0	VP19c	Capsid protein; forms triplexes together with UL18	V (C)
UL39	46470-48977	835	91.1	RR1	Nucleotide synthesis; large subunit of ribonucleotide reductase	NS
UL40	48987-49898	303	34.4	RR2	Nucleotide synthesis; small subunit of ribonucleotide reductase	NS
UL41	51498-50401 r	365	40.1	VHS	Gene regulation (inhibitor of gene expression); virion host cell shutoff factor	V (T)
UL42	51628-52782	384	40.3		DNA replication; polymerase accessory subunit of UL30/UL42 complex	NS
UL43	52842-53963	373	38.1		Unknown; type III membrane protein	V (E)
UL44	54029-55468	479	51.2	gC	Viral entry (virion attachment); glycoprotein C; type I membrane protein; binds to heparan sulfate	V (E)
UL26.5	56535-55699 r	278	28.2	VP22a	Scaffold protein; substrate for UL26; required for capsid formation and maturation	pC
UL26	57273-55699 r	524	54.6	VP24	Scaffold protein; proteinase; required for capsid formation and maturation	pC
UL25	58911-57307 r	534	57.4		Capsid-associated protein; required for capsid assembly	V (C)
UL24	59519-59004 r	171	19.1		Unknown; type III membrane protein	?
UL23	59512-60474	320	35.0		Nucleotide synthesis; thymidine kinase	NS
UL22	60610-62670	686	71.9	gH	Viral entry (fusion); cell-cell spread; glycoprotein H; type I membrane protein; complexed with gL	V (E)
UL21	66065-64488 r	525	55.2		Capsid-associated protein	V (?)
UL20	66172-66657	161	16.7		Viral egress; type III membrane protein	?
UL19	66744-70736	1,330	146.0	VP5	Major capsid protein; forms hexons und pentons	V (C)
UL18	70896-71783	295	31.6	VP23	Capsid protein; forms triplexes together with UL38	V (C)
UL15 (Ex2)	73115-71979 r	735	79.1		DNA cleavage-encapsidation; terminase subunit; interacts with UL33, UL28, and UL6	pC
UL15 (Ex1)	77065-75995 r
UL17	73166-74959	597	64.2		DNA cleavage-encapsidation	V (T)
UL16	74986-75972	328	34.8		Unknown	?
UL14	77064-77543	159	17.9		Unknown	?
UL13	77513-78709	398	41.1	VP18.8	Protein-serine/threonine kinase	V (T)
UL12	78675-80126	483	51.3		DNA recombination; alkaline exonuclease	?
UL11	80084-80275	63	7.0		Viral egress (secondary envelopment); membrane-associated tegument protein	V (T)/PICK>
UL10	81935-80754 r	393	41.5	gM	Viral egress (secondary envelopment); glycoprotein M; type III membrane protein; C terminus interacts with UL49; inhibits membrane fusion in transient assays; complexed with gN	V (E)
UL9	81934-84465	843	90.5	OBP	Sequence-specific ori-binding protein	NS
UL8.5	83053-844465	470	51.0	OPBC	C-terminal domain of UL9	?
UL8	84462-86513	683	71.2		DNA replication; part of UL5/UL8/UL52 helicase-primase complex	NS
UL7	87479-86679 r	266	29.0		Unknown	?
UL6	89301-87370 r	643	70.3		Capsid protein; portal protein; docking site for terminase	V (C)
UL5	89300-91804	834	92.1		DNA replication; part of UL5/UL8/UL52 helicase-primase complex; helicase motif	NS
UL4	91863-92300	145	15.8		Nuclear protein	?
UL3.5	93150-92476 r	224	24.0		Viral egress (secondary envelopment); membrane-associated protein	?
UL3	93860-93147 r	237	25.6		Nuclear protein	NS
UL2	94866-93916 r	316	33.0	UNG	Uracil-DNA glycosylase	NS
UL1	95314-94844 r	156	16.5	gL	Viral entry; cell-cell spread; glycoprotein L; membrane-anchored via complex with gH	V (E)
EP0	97713-96481 r	410	43.8	ICP0	Gene regulation (transactivator of viral and cellular genes); early protein	?
IE180 (IRS)	107511-103171r	1,446	148.6	ICP4	Gene regulation; immediate early protein	NS
IE180 (TRS)	137091-141431
US1 (IRS)	115995-117089	364	39.6	RSp40/ICP22	Gene regulation	?
US1 (TRS)	128607-127513 r
US3 (minor)	118170-119336	388	42.9	PK	Minor form of protein kinase (53-kDa mobility)	?
US3 (major)	118332-119336	334	36.9	PK	Viral egress (nuclear egress); major form of protein kinase (41-kDa mobility)	V (T)
US4	119396-120892	498	53.7	gG	Glycoprotein G (secreted)	secreted
US6	121075-122277	400	44.3	gD	Viral entry (cellular receptor binding protein); glycoprotein D; type I membrane protein	V (E)
US7	122298-123398	366	38.7	gI	Cell-cell spread; glycoprotein I; type I membrane protein; complexed with gE	V (E)
US8	123502-125235	577	62.4	gE	Cell-cell spread; glycoprotein E; type I membrane protein; complexed with gI; C terminus interacts with UL49	V (E)
US9	125293-125589	98	10.6	11K	Protein sorting in axons; type II tail-anchored membrane protein	V (E)
US2	125811-126581	256	27.7	28K	Unknown	?

^aNumbering starts at +1 on the U_L end of the genome. r indicates ORF encoded on reverse strand.

^bFunction or property as demonstrated for the PRV and/or HSV-1 homolog.

^cV (C), virion capsid component; V (T), virion tegument component; V (E), virion envelope component; V (?), virion component of unknown subviral localization; pV, primary enveloped virion precursor component (not found in mature virion); NS, nonstructural protein; pC, present in intranuclear capsid precursor forms but not found in mature virion; ?, unknown.

Table 3 lists the known PRV genes, including PRV homologs of HSV-1 genes, and summarizes the characteristics of the gene products. There are 72 ORFs predicted to encode 70 different proteins, as the genes encoding the IE180 and US1 protein are found twice, once in the internal repeat sequence (IRS) and once in the terminal repeat sequence (TRS). Distinct functions had been ascribed to the two US3 protein forms encoded by the major and minor US3 transcripts (73). Consequently, each form was considered to be encoded by a distinct ORF. In contrast, the ORFs contained in the major and minor UL37 transcripts were counted as a single ORF. All ORF start locations were assumed to be the first possible ATG, unless demonstrated otherwise. Nearly half the gene products can be found or are presumed to be in the mature virion (31 out of 72 ORFs, with 15 unknown). The properties and functions assigned to each gene produced were based on the studies of the PRV proteins and/or HSV-1 homologs. A more detailed description of what is known about PRV and HSV-1 is available at the Los Alamos sequence database (see Materials and Methods). A significant number of genes have not been assigned any clear function yet (20 out of 72 ORFs), though it is possible that some of these genes play a strictly structural role in the virion envelope or tegument.

As concerns ORF-1, experimental data indicate that there is an upstream in-frame extension, designated ORF1.2, with probable start codons at positions 1252 or 1375 (unpublished data). All but three PRV genes (ORF-1, ORF1.2, and UL3.5) have homologs in HSV-1. ORF-1 and ORF1.2 are located at the left terminus of the PRV U_L region and show only homology to the first ORF of EHV-1 strain Ab4 (3). UL3.5 is conserved in many alphaherpesviruses, including BHV-1, EHV-1, VZV, ILTV, and MDV, but not HSV-1 or HSV-2. In marked contrast, a number of HSV-1 genes do not seem to have a PRV counterpart: US5 (gJ), US8.5, US10, US11, US12, γ₁34.5, ORF P, ORF O, UL9.5, UL10.5, UL20.5, UL27.5, UL43.5, UL45, UL55, and UL56 (63).

Systematic search for core elements of gene expression control.

Initially, all available DNA sequences were examined for their annotated information. While this approach yielded a complete and consistent annotation of ORFs, it failed to provide a complete picture of transcriptional elements and DNA repeats. We therefore took a systematic approach to search for these elements. Most, if not all, genes in the HSV-1 genome are transcribed as capped and polyadenylated mRNAs by host RNA polymerase II (64). It is widely assumed that the homologous genes in PRV are similarly transcribed. Computer prediction programs were used to identify RNA polymerase II transcriptional control elements, including core promoters, splice sites, and polyadenylation sites. A visual search for short repeat elements was also performed.

Search for transcription polyadenylation signals.

Two sequence elements make up the core of mammalian 3′ mRNA processing signals directing mRNA cleavage and polyadenylation. The first element, located 10 to 30 bases upstream of the cleavage site, is the conserved poly(A) signal AAUAAA and is found in 90% of all sequenced polyadenylation sites. In the remaining 10%, the sequence found differs only by a single substitution, with AUUAAA the most common variant. The second element is the downstream element (DE), a U- or GU-rich sequence located 20 to 40 bases after the cleavage site (reviewed in reference 16).

The PolyADQ program was used to search for all potential polyadenylation signals in the PRV genome. This program was designed to detect and evaluate potential poly(A) signals in human DNA sequences using weight matrices for base composition and position in the DE (69). Table 4 lists the results by gene along with an associated score between 0 and 1 that primarily reflects the presence of a consensus DE. The table lists the genes directly upstream of the poly(A) signals and the length of the predicted 3′ UTR.

TABLE 4.

PRV polyadenylation signals predicted by PolyADQ

Gene	Sequence	Locationa	Score	3′ UTR sizeb		Evidencec (reference[s])
				Pred.	Exptl
ORF1.2	AATAAA	2262-2267	0.735	28		Ka type 4 (3)
ORF-1				28		Ka type 4 (3)
UL54	AATAAA	2730-2725 r	0.525	25		Ka type 4 (3)
UL53				1,190		Ka type 4 (3)
UL52				2,083		Ka type 4 (3)
UL51	AATAAA	8433-8438	0.222	85		Ka type 4 (3)
UL50	AATAAA	8448-8443 r	0.333	104
Orphan	AATAAA	9015-9020	0.031	NA		—d
UL49.5	AATAAA	10336-10341	0.455	808		Ka type 3 (28)
UL49				21		Ka type 3 (28)
UL48	AATAAA	16153-16158	0.817	4,533		Ka type 3 (7, 28)
UL47				2,180		Ka type 3 (7)
UL46				80	78	TNL type 1 (35), Ka type 2 (7)
UL27	AATAAA	16835-16830 r	0.322	44	46	TNL type 1 (36), Ka type 3 (7)
UL28				2,656
UL29	AATAAA	21741-21736 r	0.601	72	74	TNL type 1 (35)
UL30	AATAAA	28768-28773	0.274	41
UL31	AATAAA	28620-28615 r	0.693	78	74	TNL type 1 (35)
UL32				886
UL33	AATAAA	32565-32570	0.537	1,351
UL34				404		Ka type 3 (27)
UL35				38
Orphan	AATAAA	33047-33052	0.031	NA		—d
UL36	AATAAA	33061-33056 r	0.872	24
UL37 (M)g	AATAAA	42356-42351 r	0.395	22		Ka type 3 (8)
UL37 (m)g				22		Ka type 3 (8)
UL38	AATAAA	46325-46330	0.180	76		Ka type 3 (8)
UL39	AATAAA	49959-49964	0.621	1,007		NIA-3 type 4 (22)
UL40				86		NIA-3 type 4 (22)
UL41	AATAAA	50405-50400 r	0.382	21
Orphan	AATAAA	52778-52773 r	0.265	NA
UL42	AATAAA	52781-52786	0.331	24	18	TNL type 1 (36), NIA-3 type 4 (22)
UL43	ATTAAA	53946-53951	0.007	8	25	TNL type 1 (36), NIA-3 type 4 (22)
UL44	AATAAA	55500-55505	0.496	57	21 ± 40	Be type 2, 4 (60), NIA-3 type 4 (22)
Orphan	AATAAA	55608-55613	0.178	NA
UL26.5	AATAAA	55659-55654 r	0.707	65	48 ± 40	NIA-3 type 2, 4 (10), Ka type 3 (23)
UL26				65	48 ± 40	NIA-3 type 2, 4 (10), Ka type 3 (23)
UL25				1,673	1,656 ± 40	NIA-3 type 2, 4 (10), Ka type 3 (23)
UL24				3,370		Ka type 3 (23)
UL23	AATAAA	60585-60590	0.438	136		Ka type 4 (42)
UL22	AATAAA	62652-62657	0.268	7		Ka type 4 (42)
Orphan	AATAAA	63531-63526 r	0.382	NA
UL21	AATAAA	64502-64497 r	0.064	11		NIA-3 type 4 (22)
UL20	ATTAAA	66653-66658	0.005	21
UL19 (M)	ATATAAA	70838-70844	—e	128	121	Indiana S type 1, 4 (77)
UL19 (m)	AATAAA	71841-71846	0.449	1,130		Indiana S type 4 (77)
UL18				83
UL17	AATAAA	75957-75962	0.381	1,023
UL16				10
UL15	AATAAA	71897-71892 r	0.663	107
Orphan	AATAAA	78325-78320r	0.018	NA		—d
UL14	AATAAA	80280-80285	0.874	2,762		NIA-3 type 4 (22)
UL13				1,596		NIA-3 type 4 (22)
UL12				179	180	TNL type 1 (37), NIA-3 type 4 (22)
UL11				30
UL10	AATAAA	80758-80753 r	0.475	21		Ka type 3 (24)
UL9	AATAAA	86516-86521	0.797	2,076		Ka type 4 (24)
UL8.5				2,076		Ka type 4 (24)
UL8				28		Ka type 4 (24)
UL7	AATAAA	86532-86527 r	0.098	172		Ka type 4 (24)
UL6				863		Ka type 4 (24)
UL5	AATAAA	91895-91900	0.478	116		Not functionald,f
UL5	AATAAA	92394-92399	0.109	615		In-Fh type 4 (21)
UL4				119		In-Fh type 4 (21)
UL3.5	AATAAA	92475-92480	0.058	21		In-Fh type 3 (20)
UL3				692		In-Fh type 3 (20)/PICK>
UL2				1,461		In-Fh type 3 (20)
UL1				2,389		In-Fh type 3 (20)
EP0	AATAAA	96273-96268 r	0.559	233	234	In-Fh type 1, 4 (13)
LLT (IRS)	AATAAA	109092-109097	0.665	NA		Be type 1, 4 (13)
Orphan (TRS)	AATAAA	135510-135505 r	0.665	NA
IE180 (IRS)	AATAAA	102719-102714 r	0.672	477	475	In-Fh type 1, 4 (14), Ka type 2 (11)
IE180 (TRS)	AATAAA	141883-141888	0.672	477	475	In-Fh type 1, 4 (15), Ka type 2 (11)
US1 (IRS)	AATAAA	117191-117196	0.560	126	125	TNL type 1 (36), Ka type 3 (27)
US1 (TRS)	AATAAA	127411-127406 r	0.560	126	125	TNL type 1 (36), Ka type 3 (27)
Orphan (IRS)	AATAAA	117733-117728 r	0.424	NA
Orphan (TRS)	AATAAA	126869-126874	0.424	NA
Orphan	AATAAA	118239-118244	0.717	NA
US3 (M)	AATAAA	120951-120956	0.431	1,580		Ka type 3 (78), NIA-3 type 3 (74)
US3 (m)
US4				84		Ka type 3 (78), NIA-3 type 4 (74)
US6	AATAAA	123394-123399	0.673	1,142	1,105 ± 90	In-Fh type 2 (43)
US7				21
US8	AATAAA	125697-125702	0.189	487	497	TNL type 1 (35)
US9				133	143	TNL type 1 (35)
US2	AATAAA	126628-126633	0.452	72	59 ± 12	In-Fh type 2 (43), NIA-3 type 3 (74)

^aNumbering starts at +1 on the U_L end of the genome. r indicates reverse strand direction.

^bPredicted (Pred.) polyadenylation sites were set at 20 bases downstream of the poly(A) signal sequence. Exptl, experimental; NA, not applicable, no protein-coding RNA.

^cEvidence abbreviations: type 1, 3′ cDNA sequence; type 2, S1 mapping; type 3, mRNA size and sense; type 4, mRNA size only.

^dNot included in annotated sequence.

^eExperimentally determined, not found as a predicted poly(A) site.

^fContradicted by UL5 mRNA size and UL4 TSS.

^gM, major transcript; m, minor transcript.

Northern blot analyses, S1 nuclease transcript mapping, and cDNA nucleotide sequence information were used to assess the functional significance of the predicted poly(A) signals. Sequenced 3′ ends of cDNAs allow the precise determination of the poly(A) cleavage site and of the 3′ UTR length. Transcript mapping with S1 nuclease allows a less precise mapping of the 3′ ends of mRNAs and of the 3′ UTR length. In all 21 cases, the predicted and measured 3′ UTR lengths were nearly identical, validating the assignment of poly(A) signals to genes immediately upstream. Depending on the probes used, Northern blot analyses can define the location and orientation of mRNAs, or at the very least provide an estimate of mRNA sizes. Northern blot information has also been used to demonstrate the existence of 3′ coterminal transcripts. Given the relative rarity of poly(A) signals in the PRV genome due to the G+C-rich nature, an mRNA size estimate alone can lend reasonable support for the functional usage of a given poly(A) site. In all but one case, further detailed below, the mRNA sizes were consistent with our poly(A) signal assignment, and the experimentally determined mRNA sizes are listed below in Table 6.

TABLE 6.

PRV core promoters predicted by neural networkⁱ

Gene	Promoter score	TATA sequence	TATA locationa	TSS location		mRNA size (kb)		5′ UTR Calc.	Kozak (of 13)c	TSS evidenced	Note
				Predicted	Exptl	Calc.b	Exptl
ORF1,2	NP	NP	NP	NP	ND	NP	1.4-1.8e	NP	6		In-frame ATG (7/13), +123 nt
ORF-1	1.00	TATTAAC	1584-1590	1616	ND	>0.67	0.6c	20	8
UL54	1.00	GTTAAAG	3873-3867 r	3841 r	ND	>1.14	1.6c	26	9
UL53	0.93	TACAAAG	4890-4884 r	4859 r	ND	>2.20	2.8c	23	8
UL52	0.97	CTCATAA	7722-7716 r	7690 r	ND	>4.98	5.6c	14	6		In-frame ATG (8/13) 12 nt later
UL51	0.99	GCTAAAA	7492-7498	7522	ND	>0.94	1.3c	141	6		In-frame ATG (8/13) 141 nt later
UL50	0.99	TATAAAA	9458-9452 r	9427 r	ND	>1.01	ND	94	8		Within 100 bp of UL49 TATA
UL49.5	1.00	AATAAAA	9015-9021	9046	ND	>1.32	1.3	211	8		Distal TATA
UL49.5	1.00	TATAAAA	9174-9180	9204	ND	>1.16	1.3	53	8		Proximal TATA, best fit for mRNA size
UL49	1.00	TATAAAG	9549-9555	9579	ND	>0.78	0.9	12	8		Within 100 bp of UL50 TATA
UL48	1.00	TATAAAT	10332-10338	10366	ND	>5.81	5.6-6.0	38	6		Bifunctional TATA-poly(A) signal, in-frame ATG (9/13), at +168 nt
UL47	1.00	TATAAAG	11695-11701	11727	ND	>4.45	4.5	19	10
UL46	0.99	CATTTAT	13963-13969	13994	ND	>2.18	2.4	23	11
UL27	0.96	GATATAT	19722-19716 r	19691 r	ND	>2.88	3.1	96	7
UL28	0.97	AATAAAG	21741-21735 r	21713 r	ND	>4.91	ND	73	10		Bifunctional TATA-poly(A) signal
UL29	1.00	TTTAAGA	25518-25512 r	25488 r	ND	>3.78	ND	172	10		Bidirectional TATA (UL29, UL30)
UL30	0.99	TCTTAAA	25512-25518	25544	ND	>3.18	ND	62	7		Bidirectional TATA (UL29, UL30)
UL31	0.91	TATTTAA	29614-29608 r	29584 r	ND	>0.99	ND	96	8
UL32	1.00	TTTATAG	31240-31234 r	31212 r	ND	>2.61	ND	319	9		Bidirectional TATA (UL32, UL34)
UL33	NP	NP	NP	NP	ND	NP	ND	NP	7
UL34	1.00	TATAAAG	31235-31241	31275	ND	>1.33	1.3	132	8		Bidirectional TATA (UL32, UL34)
UL35	1.00	GTTAAAG	32182-32188	32213	ND	>0.38	ND	28	8
UL36	NP	NP	NP	NP	ND	NP	ND	NP	7
UL37 (M)	0.99	TATAATG	45115-45109 r	45087 r	45099+/−10 r	>2.76	3.5-4.0	57	7	Ka type 1 (8)	TSS match, bidirectional TATA (UL37 [M], UL38), aa 28-919
UL37 (m)	1.00	TATAAGG	45251-45245 r	45221 r	45224+/−10 r	>2.89	3.5-4.0	110	6	Ka type 1 (8)	TSS match, in-frame ATG (7/13), at +81 nt
UL38	1.00	TATAAGA	45112-45118	45140	45145+/−6	>1.21	1.1	28	9	Ka type 1 (8)	TSS match, bidirectional TATA (UL37 [M], UL38)
UL39	1.00	AATAAAA	46325-46330	46356	ND	>3.63	3.7e	112	6		Bifunctional TATA-polyA signal, in-frame ATG (11/13), at +135 nt
UL39	0.99	GCTAAAA	46538-46544	46569	ND	>3.46	3.7e	37	11		Amino acids 46 to 835
UL40	0.98	CATATAA	48868-48874	48900	ND	>1.08	1.2e	87	10
UL41	0.99	CTTATAT	51571-51565 r	51541 r	ND	>1.16	ND	43	8		Bidirectional TATA (UL41, proximal UL42)
UL42	0.99	TCTAAAA	51526-51532	51556	ND	>1.25	1.5e	72	10		Distal TATA
UL42	1.00	CATATAA	51564-51570	51596	ND	>1.21	1.5e	32	10		Proximal TATA, bidirectional TATA (UL41, UL42)
UL43	1.00	TATAAAA	52798-52804	52829	ND	>1.16	1.6c	13	10
UL44	1.00	TTTAAAA	53986-53992	54016	54018+/−5	>1.51	1.6-1.7e	13	10	Be type 1 (61)	TSS match
UL26.5	0.99	CCTAAAA	56577-56571 r	56545 r	56546+/−1 r	>0.91	1.0	10	9	NIA-3 type 2 (10)	TSS match, in-frame ATG (11/13), at +132 nt
UL26	0.81	TATATCC	57333-57327 r	57304 r	57400+/−1 r	>1.67	1.7	31	13	NIA-3 type 3 (10)	TSS mismatch
UL25	0.98	GATAAGG	58956-58950 r	58927 r	58963+/−1 r	>3.29	3.4	16	10	NIA-3 type 3 (10)	TSS mismatch, published TSS predicts a UL25 5 amino acids longer
UL24	0.99	CGTAAAT	59665-59659 r	59633 r	ND	>4.00	3.9	114	5
UL23	NP	NP	NP	NP	ND	NP	1.5e	NP	8
UL22	0.83	TATAAAG	60560-60566	60590	60589+/−1	>2.09	2.3e	20	10	Ka type 3h	TSS match
UL21	0.98	TTTAAAG	66126-66120 r	66097 r	ND	>1.62	1.8e	32	12		Bidirectional TATA (UL20, UL21)
UL20	1.00	TTTAAAC	66121-66127	66151	ND	>0.53	ND	21	9		Bidirectional TATA (UL20, UL21)
UL19	1.00	CATTAAA	66652-66658	66683	66684+/−1	>4.18f	4.4e,f	61	8	Ind S type 3 (77)	TSS match, bifunctional TATA-poly(A) signal
UL18	1.00	TATATAA	70837-70843	70867	ND	>1.00	ND	29	11		Bifunctional TATA-poly(A) signal
UL17	0.99	TATAAAG	73062-73068	73092	ND	>2.89	ND	74	11
UL16	1.00	TATAAGG	74704-74710	74734	ND	>1.25	ND	252	10
UL15	0.99	CATAAAG	77204-77198 r	77213 r	ND	>2.43g	ND	108	8		Within 100 bp of UL13 TATA
UL14	1.00	TTGAAAA	76965-76971	76996	ND	>3.31	3.5e	68	6		In-frame ATG (7/13), at +138 nt
UL13	1.00	AACAAAA	77272-77278	77304	ND	>3.00	3.2e	210	9		Within 100 bp of UL15 TATA
UL12	1.00	TATTAAC	78322-78329	78351	78501+/−8	>1.95	2.1e	324	9	TNL type 3 (37)	TSS mismatch
UL11	NP	NP	NP	NP	ND	NP	ND	NP	10
UL10	0.99	TATCAGT	82219-82213 r	82189 r	ND	>1.46	1.6	254	11
UL9	0.89	TCTATCA	81883-81889	81907	ND	>4.64	5.1e	27	8
UL8.5	NP	NP	NP	NP	ND	NP	3.6c	NP	6		In-frame ATGs (6/13), at +21 nt, and (9/13), at +39 nt
UL8	0.99	TATAAAC	84322-84328	84350	ND	>2.20	2.4e	112	8
UL7	0.97	TTTAAGA	87567-87561 r	87537 r	87538+/−1 r	>1.03	1.3e	58	10	Be type 3 (51)	TSS match
UL6	0.31	GCTAATA	89241-89235 r	89208 r	89212+/−1r	>2.70	3.1e	24	9	Be type 3 (51)	Amino acids 40 to 643, low-scoring TATA, bidirectional TATA (UL5,UL6)
UL5	0.86	TATTAGC	89235-89241	89257	89262+/−1	>3.16	3.6	39	11	In-Fh type 3 (21)	TSS match, low-scoring TATA, bidirectional TATA (UL5, UL6)
UL4	1.00	CATATAT	91757-91763	91789	91783+/−1	>0.64	0.9	80	7	In-Fh type 3 (21)	TSS match, in-frame ATG (9/13), +108 nt
UL4	0.99	AATAAAG	91895-91901	91925	ND	>0.50	0.9	46	9		Amino acids 37 to 145
UL3.5	0.95	CATAAAA	93362-93356 r	93332 r	ND	>0.88	0.9	182	9
UL3	0.99	TCAAAAG	94078-94072 r	94047 r	ND	>1.59	1.8	187	7
UL2	0.99	CTTAAAT	94940-94934 r	94908 r	ND	>2.45	2.7	42	8
UL1	0.97	GATAAAA	95360-95354 r	95330 r	95329+/−1 r	>2.92	3.3	16	9	Ka type 3 (40)	TSS match
EP0	1.00	TAAAAAA	97736-97730 r	97707 r	97754+/−1 r	>1.44	1.75e	123	9	In-Fh type 3 (13)	TSS mismatch, amino acids 44 to 140
LLT (IRS)	1.00	TATATAA	96079-96085	96111	96112	>8.39g	8.5e	NA	NA	Be type 3 (13)	TSS match
IE180 (IRS)	0.98	CTTATAA	107804-107798 r	107773 r	107773+/−1 r	>5.08	5.6e	262	9	In-Fh type 2 (14)	TSS match
IE180 (TRS)			136798-136804	136829	138629+/−1					& Ka type 1 (11)	TSS match
US1 (IRS)	0.99	GATAAAG	111521-115217	115241	ND	>1.70g	1.8	478	12		5′ UTR contain 2 introns
US1 (TRS)			129391-129385 r	129361 r	ND						5′ UTR contain 2 introns
US3 (m)	0.42	GATATCG	118076-118082	118108	118108+/−13	>2.87	2.7-3.0	62	8	NIA-3 type 1 (74)
US3 (M)	0.70	AATAAAG	118239-118245	118268	118266+/−5	>2.71	2.7-3.0	64	10	NIA-3 type 1 (74)
US4	1.00	TATAAAA	119332-119338	119363	ND	>1.61	1.6	33	7
US6	1.00	CATAAAA	121024-121030	121054	121059+/−10	>2.37	ND	21	10	In-Fh type 1 (43)	TSS match
US7	0.99	GCTAAAA	122253-122259	122285	ND	>1.14	ND	13	7
US8	0.99	TTTTAAA	123452-123458	123484	123469+/−20	>2.23	ND	18	8	In-Fh type 1 (43)	TSS match
US9	1.00	CTTAAAT	125231-125237	125261	ND	>0.47	ND	32	11		Unused in-frame ATG (7/13) in 5′ UTR (9)
US2	1.00	AATAAAT	125697-125703	125726	125726+/−10	>0.93	1.2	85	10	NIA-3 type 1 (74)	TSS match, bifunctional TATA-poly(A) signal

^aNumbering starts at +1 on the U_L end of the genome; r indicates reverse strand direction.

^bmRNA size predicted does not include poly(A) tail (usually 150 to 300 bp). Refer to Table 4 for mRNA size references.

^cIdentity of the first ATG to the Kozak consensus, GCCGCCRCCATGG.

^dEvidence abbreviations: type 1, S1 mapping; type 2, primer extension with 2 primers; type 3, primer extension with 1 primer.

^eSense of mRNA not established.

^fmRNA size was calculated using experimental poly(A) site at nt 70857. If the the poly(A) signal at nt 71841 to 71846 was used, the predicted mRNA was >5.18 kb, associated with a minor transcript observed at 5.5 kb (77).

^gCalculated mRNA size after splicing.

^hUL22 evidence is unpublished (B. Klupp and T. C. Mettenleiter).

ⁱAbbreviations: ND, not determined; NP, no prediction; NA, not applicable, no protein-coding RNA; M, major transcript; m, minor transcript.

Since different weight matrices were used to assign scores to AATAAA-based or ATTAAA-based signals, their scores cannot be compared. However, within a given type of poly(A) signal, the experimental results were used to assign a minimal cutoff score for poly(A) signals to be included in the annotated genome. For the common AATAAA signals, experimental support was obtained for a score as low as 0.058 (UL1, UL2, UL3, and UL3.5 coterminal transcripts). We thus set the cutoff at 0.05 for maximal sensitivity, resulting in the elimination of 3 of the 10 poly(A) signals with no known upstream genes (so-called orphan signals). For the two extremely low-scoring ATTAAA signals, experimental support was obtained from the UL43 cDNA sequence, and both signals were included in the annotation. Table 4 also indicates that 65% of known PRV genes (48 of 73) are predicted to share coterminal transcripts with another gene or with as many as three other genes. All available experimental data listed support this prediction.

Search for splice sites.

Splicing of mRNA involves the recognition of acceptor and donor sequences by the spliceosome. We searched for splice donor and acceptor sites in PRV genes by using a neural network splice site prediction program conditioned for human splice site recognition. Sequences from cDNA had established the existence of three introns in PRV so far: two in the 5′ UTR of US1 (27) and one in the LLT (13). A stringent search of the entire PRV genome found only one splice donor-acceptor pair in all the predicted PRV transcripts, matching the coordinates of the second intron in the 5′ UTR of US1. The search failed to accurately predict the other two known introns and a putative PRV intron in UL15, a homolog of the spliced UL15 gene of HSV-1. UL15 is made up of two exons and is well conserved among herpesviruses. PRV and HSV-1 UL15 possess similar exon lengths, strong protein sequence homology, and a good DNA sequence homology at the donor and acceptor sites. The DNA sequences of splice donors and acceptors for PRV UL15, US1, and LLT compare favorably to the eukaryotic consensus (Fig. 1). Remarkably, the predicted UL15 splice donor site (PRV Ka) does not contain the invariant GT dinucleotide at the start of the intron. Whether this predicted donor site is really functional remains to be determined, but it is worth noting that identical splice sequences were found for UL15 in the Ea strain (GenBank accession no. AY189899), a recent PRV isolate from Wuhan (China) (12).

FIG. 1.

Comparison of the PRV splice donor and acceptor site sequences and the mammalian consensus. Intron sequences are underlined, and the locations of the sites in the PRV genome are indicated. For US1, the site locations given are for the IRS gene copy. For the TRS copy of US1, the sites are located as follows: US1 intron 1 donor (129209 to 129201) and acceptor (129103 to 129085); US1 intron 2 donor (129035 to 129027) and acceptor (128895 to 128877). M = A or C; R = A or G; Y = C or T.

Search for repeat elements.

The PRV genome carries a variety of repeated DNA sequences. Seventeen different direct repeat regions were found: 11 in the U_L segment, 1 in the U_S segment, and 6 each in the IRS and the TRS (Table 5 and Fig. 2). The IRS and TRS themselves are large inverted repeats. The location of the repeats suggests a possible role in transcriptional insulation: 5 of the 11 U_L direct repeat regions and 2 of the 6 IRS/TRS repeat regions were found located between two poly(A) signals from convergent transcripts. The repeats may serve to prevent any accidental read-through by RNA polymerase into the oppositely transcribed gene. Furthermore, three direct repeat regions and two inverted repeats were found in the first kilobase of the linear genome at the U_L terminus, while two of the five repeat regions in the TRS were found in the last 1,000 bp of the linear genome. These repeat regions may serve to insulate against read-through transcription after genome circularization during latency in neurons. Alternatively, they could play a role in the process of genome circularization itself.

TABLE 5.

DNA repeats

Locationa	Repeat		Type
	Unit	No.
3-84	82-mer	1	Inverted repeat of nt 442 to 523, Ka
156-251	28-mer	3	Imperfect spaced direct repeats, Ka
442-523	82-mer	1	Inverted repeat of nt 3 to 84, Ka
529-655	40-mer	3	Imperfect spaced direct 36-, 38- & 40-mer repeat, Ka
751-958	26-mer	8	Consecutive direct repeats, Ka
2320-2676	21-mer	17	Consecutive direct repeats, Kab
16218-16802	15-mer	39	Consecutive direct repeats, Kab
32680-32881	10-mer	21	Imperfect consecutive direct repeats (8 to 11 mers), Kab
50181-50268	11-mer	8	Consecutive direct repeats, Kab
63110-63319	15-mer	14	Consecutive direct repeats, near OriL, Ka
63388-63453	11-mer	6	Consecutive direct repeats, near OriL, Ka
80326-80541	12-mer	18	Consecutive direct repeats, Kab
95518-95616	11-mer	9	Consecutive direct repeats, Ka
101141-117942	IRS	1	Inverted repeat of TRS, 16.8 kb, separates UL and US regions, Ka
101376-101501	22-mer	5.5	Consecutive direct repeats, Ka
101650-101775	63-mer	2	Consecutive direct repeats, Ka
108494-108683	19-mer	10	Consecutive direct repeats, Ka
114359-115158	280-mer	2.8	Imperfect consecutive direct repeats forming the OriS, Ka
117279-117687	35-mer	11.5	Consecutive direct repeats, Ka
117752-117841	10-mer	9	Consecutive direct repeats, Ka
120218-120378	50-mer	2	Spaced direct repeats, in US4 CDS, Rice
126660-143461	TRS	1	Inverted repeat of IRS, 16.8 kb, end of linear genome, Ka
126761-126850	10-mer	9	Consecutive direct repeats, Kab
126915-127323	35-mer	11.5	Consecutive direct repeats, Kab
129444-130243	280-mer	2.8	Imperfect consecutive direct repeats forming the OriS, Ka
135919-136108	19-mer	10	Consecutive direct repeats, Ka
142827-142952	79-mer	2	Overlapping direct repeats, Ka
143101-143312	22-mer	5.5	Overlapping direct repeats, Ka

^aNumbering starts at +1 on the U_L end of the genome.

^bSeparates the poly(A) signals of two converging transcripts.

Search for promoters.

The core promoters responsible for mRNA initiation exhibit considerable diversity. Nonetheless, four sequence elements showing some conservation in sequence and location are frequently found: the TATA box, the initiator element (Inr), the downstream promoter element (DPE), and the TFIIB recognition element (BRE) (reviewed in reference 68). The TATA box is a short sequence (TATAAAA) frequently found 25 to 30 bp upstream of the TSS that binds the TATA-binding protein (TBP), a subunit of the TFIID transcription factor complex. The less-well-defined Inr (PyPyAN[T/A]PyPy) often overlaps the TSS, binds components of the TFIID complex, and can act alone or synergistically with a proximal TATA box to enhance transcription initiation. The DPE is a 5-bp element that is sometimes found 28 nucleotides downstream of the TSS and functions with an Inr to bind TFIID. Finally, the recently discovered BRE is a 7-bp motif that serves to bind basal transcription factor TFIIB and is located just upstream of the TATA box.

A human core promoter prediction was used for an initial high-stringency search of the entire PRV genome, finding core promoters for 47 of the 73 genes. A search for the nearest consensus to a TATA box in these promoters was performed, and it found them all located 34 to 29 bp upstream of the predicted TSS. To find promoters for the remaining 26 genes, the search parameters were relaxed and the upstream 350 bp of all ORFs were analyzed, yielding promoters and TATA box predictions for all but 6 of the 73 genes. The genes regulated by a given promoter were defined by examining the translation product of the predicted transcripts. Table 6 lists the results by genes, along with the TATA and TSS locations and the associated promoter score (between 0 and 1). Unless performed at very high stringency, the searches often identified more than one putative promoter for a given ORF, in close proximity to each other. As such, the experimentally measured mRNA sizes, with their low precision and only rough estimation of the size of poly(A) tails, were of no help in validating our particular promoter predictions. In contrast, S1 nuclease transcript mapping and primer extension data can accurately assess the 5′ end of transcripts (TSS location), and they provided a useful test for the validity of our promoter predictions. The predicted and experimentally determined TSS locations and mRNA sizes are indicated in Table 6. Predicted mRNA sizes relied on the data in Tables 4 and 6, while the 5′ UTR length was calculated using the predicted location of the TSS. Table 6 describes the experimental evidence that located the TSS for 23 PRV genes, with our predicted TSS locations matching 19 of the 23. The degree of DNA identity between the sequences surrounding each ORF's start and the Kozak consensus is also indicated.

Overall genome structure and control of gene expression.

Figure 2 is a visual summary of data contained in Tables 3, 4, 5, and 6, depicting the arrangement of the 73 genes (72 ORFs and the LLT) and their predicted transcripts in the PRV genome. The genome is organized in a U_L region of 101.1 kb and a U_S region of 8.7 kb. The U_S region is bracketed by the IRS and TRS, two large inverted repeats 16.8 kb in length. Since the U_L region is not flanked by inverted repeats, the PRV genome exhibits the typical D class herpesvirus genome structure also found in VZV, BHV-1, EHV-1, EHV-4, and ILTV (62). The gene content and arrangement in the PRV genome are similar to those of HSV-1 and the other alphaherpesviruses. Indeed, the PRV genome is colinear with these viruses except for an internal inversion of 39 kb extending from UL27 (gB) to UL44 (gC) (5, 7, 23). A similar inversion is also present in the genome of ILTV, extending from UL22 to UL44 (79).

A large portion of the genome (over 83%) serves as template for transcripts. The abundance of coterminal transcripts (48 of 73 genes) was readily apparent. Seven of the 11 repeat regions could be seen separating convergent transcripts, while one set of convergent transcripts was predicted to overlap (∼194 bases) at their 3′ end (UL30/UL31). Divergent transcripts in close proximity to each other were observed 13 times. Divergent transcripts were predicted to have short overlaps at their 5′ end in five cases (∼82 to 282 bases), raising the possibility of mutual negative regulation: an increase in the transcription of one gene would reduce the transcription of the other. Nonoverlapping divergent transcripts occurred in the eight other cases, with six cases sharing the same TATA box (bidirectional TATA, noted in Table 6). In two other cases, the TATA elements were within 100 bp of each other and the genes may be coregulated by the same regulatory factors bound in proximity (noted in Table 6). All six bifunctional TATA-poly(A) sites (Table 6) resulted in the appearance of transcripts arranged in a head-to-tail fashion. Finally, completely overlapping genes transcribed in opposite orientations were seen in four cases (IE180/LLT, EP0/LLT, UL15/UL16, and UL15/UL17). Simultaneous transcription of both strands seems unlikely in some cases, as the genes are predicted to be expressed at different times and in different tissues. LLT is only expressed in latently infected neurons, while IE180 and EP0 are expressed early during productive infection. The timing of UL15 gene expression may well overlap with that of UL16 and UL17, since all three homologous HSV-1 proteins are believed to be involved in the same process of capsid maturation and assembly later in infection.

Origins of replication.

Figure 2 shows the three well-defined origins of replication found in PRV: OriL, located between UL21 and UL22 (76), and OriS, located in the IRS and TRS upstream of US1 (27). OriL and OriS contain the same sequence features: two inverted copies of the UL9 (OBP) binding sequence (GTTCGCAC) separated by a 43-bp AT-rich spacer sequence (76% A+T) (27, 41). This basic arrangement was present once in OriL and was found as three imperfect repeats in OriS, and it is very similar to the palindromic arrangement described for HSV-1 OriL and OriS (63).

An additional origin of replication had previously been proposed to be located in the BamHI-14′ fragment, the 1.3-kb terminal end of the PRV U_L region (76). However, our sequence analysis found only one UL9 consensus binding sequence in this region, at position 1243 to 1250. The PRV genome contains two more single UL9 protein recognition sequences, at positions 25580 to 25587 and 34847 to 34854. None of the three is adjacent to an AT-rich stretch of DNA. Therefore, it is questionable whether any of these has the potential to function as an origin of replication.

DISCUSSION

We report here the first complete DNA sequence for the PRV genome, fully annotated for features related to DNA (repeat elements and origins of replication), proteins (coding sequence locations, protein function and location, and signal sequences) and gene expression (locations of mRNA and transcriptional control elements). These annotations combine the results obtained by systematic searches using prediction software and carefully scrutinized experimental data from the published body of literature.

Survey of PRV genome sequence and gene content.

The genome sequence data were assembled from the sequence fragments available in the GenBank database and completed by sequencing of the remaining gaps. While the completed sequence was derived from more than one strain source (Table 2), a DNA sequence analysis showed the PRV strains to be closely related (Table 1). An evaluation of the gene content of PRV found ORF1.2 as an additional ORF to those described in reference 47, though the complete coding sequences for UL15, UL16, and UL17 were unavailable at that time. The PRV genome is thus proposed to encode one LLT and 72 genes that encode 70 different proteins (Table 3). The genes encoding the US1 and IE180 proteins are present twice, once in the IRS and once in the TRS. The major and minor forms of US3 are treated as separate genes with distinct functions (73).

While the search for new PRV protein-coding genes found no convincing candidates, it is possible that PRV contains additional genes. We found 10 poly(A) orphan signals and discarded 3 of them because of extremely low scores. A significant number of promoters not assigned to known PRV genes were also found, even at the highest-stringency search. However, the predicted translation products of these putative transcripts tended to be small or preceded by an uncharacteristically long 5′ UTR (data not shown). Thus, it is conceivable that several of these small ORFs are expressed or that non-protein-coding transcripts exist.

Computer searches of transcriptional control elements.

We searched for transcriptional control elements in the PRV genome, including core promoters and TATA boxes, splice sites, and polyadenylation sites, using computerized prediction tools. The use of these programs relied on two assumptions: (i) that the core transcriptional elements between pigs and humans would be conserved, and (ii) that the core transcriptional elements of virus and host would be very similar.

poly(A) signals.

Our poly(A) signal assignment to upstream genes implied that most or all poly(A) signals had been found and that the poly(A) signals found were all functional. We further assumed that focusing on the common consensus signals AAUAAA and AUUAAA would be sufficient, even though other variations, while rare, are known to exist (16). The experimental data in Table 4 supported these assumptions with the following two exceptions: (i) the UL19 cDNA sequence and mRNA size strongly suggest that a UL19 transcript uses an uncommon poly(A) signal, ATATAAA (77). While this sequence motif was found three more times in the PRV genome, it never affected any of our transcript predictions. (ii) The mRNA size of UL5 and the evidence that UL5 and UL4 transcripts are coterminal invalidate a functional poly(A) signal immediately downstream of the UL5 coding sequence (top strand, nucleotides [nt] 91895 to 91900) (21). It was also noted that had this poly(A) signal been functional, it would have prevented the transcription of the full-length UL4 from our predicted promoter (Table 6), as the signal is actually located in the UL4 ORF. Dean and Cheung (21) have hypothesized that transcription from the UL4 promoter might preclude the efficient use of this poly(A) signal. This is, most likely, a unique case, since all remaining experimental data agree with our predictions.

Experimental data on transcript size or 3′ transcript location exist for 58 of the 73 PRV genes, and 56 agree with our poly(A) predictions (96% accuracy). Similar to what had been observed with HSV-1 (45), the 3′ end of mRNA predicted from the location of poly(A) consensus sequences was much more reliable than the predictions of promoters and mRNA splice sites. PRV is proposed to have 44 poly(A) sites for 73 genes, while a previous analysis in HSV-1 proposed 46 poly(A) sites for 70 genes (45). The same analysis also predicted that HSV-1 transcripts were organized as 24 singlet transcripts and 19 coterminal families, highly similar to PRV's predicted 26 singlet transcripts and 18 coterminal families.

The PolyADQ scores for the various poly(A) signals were found to have very limited predictive value, and we offer three potential explanations. First, the PolyADQ weight matrices examine the first 100 bases downstream of the poly(A) signal to gauge the presence of a consensus DE. Sequences outside this window may play an important role. Second, the weight matrices were established with a limited set of false and true poly(A) signals: 81 true and 258 false AATAAA signals and 17 true and 204 false ATTAAA signals. Finally, the weight matrices were derived from human cDNA sequences, while we are examining a genome of a porcine virus.

Promoters, TATA elements, and splice sites.

Our promoter prediction approach found 72 possible promoters for 67 of the 73 genes in PRV (Table 6). In five cases (UL49.5, UL42, UL39, UL37, and UL4), two good scoring promoters were found for each ORF. It is possible that regulatory transcription factors, DNA accessibility, or competition between the two promoters favors one over the other. Alternatively, both promoters may be used and even differentially regulated: each promoter could be used at different times during infection or function in specific cell types. The experimental evidence derived from the analysis of the 5′ end of the major (M) and minor (m) UL37 transcripts agrees with our prediction of two distinct promoters, though we could not predict their different relative strengths. In the absence of better predictive tools that take into account more than just the basic core of the promoter sequences, we are unable to resolve how these dual promoters are used.

The promoter assignment to each gene assumed (i) that the first ATG after the TSS would be used, (ii) that there would be no splicing in the 5′ UTR, with the exception of the reported case in US1 (27), (iii) that all promoters would contain a TATA-like element, and (iv) in the lower-stringency promoter search that the 5′ UTR would be smaller than 310 nt.

Except for US9, none of the promoters found contained an intervening ATG before the predicted ORFs (Table 6). A direct comparison of the DNA sequences around the first ATG and the 13 nt of the Kozak consensus showed seven or more bases to be identical at most genes. In the few cases where the identity was lower, an in-frame ATG closer to the consensus was invariably found in the next 200 nt, which may indicate an additional or the true translation start site. This is the case for US9 (9): a downstream ATG close to the Kozak consensus (11 of 13) is used instead of a more divergent ATG (7 of 13) 24 nt upstream. The predictive value of these sequence comparisons is limited by two factors. The nucleotides adjacent to the ATG are known to be more important (purine at position −3 and G at position +4; CCA/GCCATGG) for efficient translation than the rest of the Kozak consensus (44), and the secondary structure of the RNA can affect the efficiency of translation (52).

Only a few genes have been found to be spliced in alphaherpesviruses. They are usually immediate-early or latency genes, as splicing is generally inhibited late in productive infections (67). A notable exception to this general rule seems to be UL15, whose spliced mRNA can be detected late (6 h postinfection) during HSV-1 infection (17).

All but three promoters (US1, UL12, and UL32) predict 5′ UTR lengths under 300 nt. Furthermore, herpesvirus genes are generally reported to contain a 5′ UTR 30 to 300 nt long (63).

Recent database analyses of Drosophila and human core promoters had found that only 30 to 40% of the promoters contain a TATAAA consensus or a sequence with one mismatch from the consensus (reviewed in reference 68). While it is possible that some TATA-less promoters exist in PRV, we have found a TATA-like consensus in almost all core promoters by using relaxed search parameters. These TATA-like elements were invariably located 34 to 28 nt upstream of the TSS. The finding of TATA-like elements at the predicted position is biologically significant and not the result of any preprogrammed bias for TATA elements in the promoter prediction program, as the neural network was trained with a set of naturally occurring core promoter sequences 51 bp long (−40 to +11 relative to the TSS).

The six genes without a predicted core promoter may either contain a long or spliced 5′ UTR, a TATA-less promoter, and/or a poorly scoring promoter. Because the human core promoter sequences used to train the program included little or no sequences downstream of the TSS, the program did not consider any contributions from the DPE. In addition to the core promoter elements, a number of highly variable sequence elements are located upstream of core promoters and serve to regulate transcription. Clearly, the prediction scores of the various promoters do not take into account the absence or presence of such variable elements or of the DPE. Still, the predictions derived from our approach have already been useful in building a near-complete map of transcripts in the PRV genome (Fig. 2).

Our predicted start sites matched the experimental data fairly well, though less data were available for the location of the 5′ end than for the 3′ end of transcripts. Three types of experimental TSS data were available: (i) primer extension data yielding a precise 5′ location but very dependent on probe location and specificity and often subject to differing interpretations of data; (ii) primer extension data with two primers, the second primer increasing data reliability; and (iii) S1 analysis, which mapped the 5′ end of transcripts with less precision but with excellent reliability. All predicted TSS matched those mapped by primer extension analysis with two primers or by S1 analysis (10 matches). The predicted TSS matched only 7 of the 11 TSS mapped by primer extensions using a single primer. The discrepancy between predicted and experimental results could not be resolved: the primers used were often located too close to or too far from the TSS to pick up our predicted TSS. Moreover, the longest extended products were always chosen as representative of the transcript start to map the TSS despite the presence of abundant extension products of smaller sizes. While the total predicted and experimental TSS locations matched 17 times out of 21, the true accuracy rate of our predictions is likely to be closer to 80% (16 of 20), since the TSS match for the two copies of IE180 was counted twice. Because the promoters for UL6 and the minor and major forms of US3 were found based on the experimental TSS locations, they are not counted as a positive match.

It had been noted in HSV-1 that TATA boxes or other promoter elements were, by themselves, of little predictive value in identifying mRNA start sites (45). Our promoter predictions were more successful, largely due to advances of the last decade: highly improved predictive core promoter programs, along with more extensive and detailed databases of known core promoters. The neural network promoter prediction is particularly useful when used in conjunction with defined parameters, such as known ORF locations and mapped TSS.

Two new core element features have been discovered by our analysis: (i) the bidirectional TATA box (occurring six times), predicted to be shared by two overlapping promoters of oppositely transcribed genes, and (ii) the bifunctional TATA-poly(A) signal, a TATA box that also serves as polyadenylation signal for a gene upstream (occurring six times). The available experimental evidence suggests that both features exist. The mapped TSS for UL37 (M) and UL38 are located 50 bp apart and closely agree with our predicted promoters and our bidirectional TATA box (Table 6). Likewise, the mapped TSS for UL5 and UL6 also agree with our predicted promoter and bidirectional TATA box (Table 6). The mapped TSS for US2 (Table 6) is 6 bp apart from the sequenced end of the US9 and US8 transcripts (Table 4), providing support for the predicted bifunctional TATA(US2)-poly(A) signal (US8/US9). Finally, two cases of divergent transcripts with TATA boxes within 100 bp of each other were also noted, which may indicate shared regulatory elements.

In bidirectional TATA boxes, the TBP may bind in either orientation to the same sequence. The binding orientation of TBP then determines at which start site the transcription preinitiation complex assembles and which of the two genes will be transcribed. In support of the idea of bidirectional TATA boxes, TBP itself has been found to bind a TATA box consensus in both orientations in solution, with only a small preference for the correct orientation. Furthermore, recent studies suggest that the dominant mechanism in determining the direction of transcription may be the activator-enhanced polarity of TBP binding (reviewed in reference 68). Bidirectional TATA boxes also suggest a simple regulatory mechanism whereby increased expression from one gene lowers the expression of the other gene.

Features of transcript architecture conserved in alphaherpesviruses.

The gene architecture is well conserved among alphaherpesviruses and can be defined by conserved blocks of genes that show homology in their protein-coding sequence and their position relative to each other. This conservation extends to details of the transcriptional architecture itself. BHV-1 is the closest known relative of PRV, and the transcription termination sites found in the annotated BHV-1 genome predict virtually the same arrangement of singlets and coterminal families that we predict for their homologs in PRV, with few exceptions. Indeed, the largest two coterminal transcript families have clearly been demonstrated to occur in both BHV-1 and PRV: UL1, UL2, UL3, and UL3.5 (20, 39) and UL24, UL25, UL26, and UL26.5 (23, 32). Similarly, the predicted HSV-1 transcript arrangement is highly homologous to the one predicted for PRV (46). As more alphaherpesviruses are examined, a picture of conserved transcriptional features is likely to emerge, including which genes are spliced, arranged in coterminal clusters or transcribed in overlapping and opposite directions. The conservation of many of these features among several viruses suggests that the transcript arrangement is critical for the viral life cycle, probably by properly regulating viral gene expression.

Significance of the transcriptional architecture for microarray analysis.

Coterminal genes and oppositely transcribed regions present a first challenge for the microarray analysis of gene expression, not just in PRV but in related alphaherpesviruses as well. The array probes commonly used are often complementary to ORFs and many are guaranteed to hybridize to different overlapping transcripts, precluding the simple assignment of signal intensity from one array spot to one gene. Mapped transcript boundaries will not only help in understanding the proper source of array spot signals but will also allow the judicious positioning of probes to regions unique to one or just a few transcripts. The high G+C content of PRV and the long 3′ UTR of many mRNAs present a second challenge, as these two factors can hinder the synthesis of labeled cDNA strands of sufficient length to encompass the ORF-based probes when oligo(dT) primers are used. The low signals for such genes are likely to be misinterpreted as indicating low expression levels. Again, knowledge of the transcription boundaries will lead to a more careful and accurate analysis.